Transformer Condition Assessment Study With Multi-Source Data Under Indoor Heat Exchange

Power transformers are critical equipment in the power grid,and once problems occur in their operation,they will not only affect the efficiency of the substation itself,but in extreme conditions may even cause huge economic losses.Power transformers that have been in operation for many years will inevitably develop various degrees of aging under the complex operating environment and a variety of external conditions,however,for actual transformers,this process change is difficult to detect through the monitoring of real-time data.In addition,the current site condition monitoring of the various data is relatively scattered,each department is divided into information and data interface can’t be interoperable,there are data is difficult to share and comprehensive use of the problem,there is no clear,the same rules for the transformer status of qualitative analysis.In the actual operation of the transformer,the transformer status may be misjudged due to the influence of the external environment on the work of the transformer such as heat dissipation in the main transformer room.In response to the above problems,the following studies are conducted in this thesis:In view of the strong limitations of the existing transformer state assessment model and the mismatch between the required indexes and the field measurement conditions,the operation mechanism and basic parameters of the transformer are analyzed,and the operation data,environmental data,thermal data,mechanical state data and test data are selected as the basis for state assessment.The method of accessing substation background data and field measurement is combined to collect the required data and prepare for the state assessment.In view of the influence of ventilation and heat dissipation in the main transformer room on oil temperature and the poor quality of transformer oil temperature data caused by deterioration of the oil temperature measuring sensor,this thesis builds a threedimensional flow field-temperature field model of the transformer based on COMSOL software and finite element analysis method to calculate the temperature field distribution of each main part of the transformer(iron core,winding and insulating oil).The effect of indoor heat dissipation in substation on oil temperature was analyzed and abnormal oil temperature data was corrected.Moreover,multi-layer perceptron neural network was used to obtain winding hot spot temperature data based on oil temperature,ambient temperature and load current inversion,which provided an important data category for transformer thermal state data and helped to improve the accuracy of subsequent state evaluation.In view of the low accuracy of transformer equipment evaluation model and the lack of effective unified evaluation criteria,the transformer load data are clustered based on k-means method and combined with the working conditions of indoor heat dissipation system of the main transformer,and different state evaluation intervals are divided.Based on the obtained data,the electrical volume,vibration amount and thermal state amount are proposed as state evaluation indexes.The sample sequence of optimal length suitable for transformer state evaluation is determined by the phasespace reconstruction method,and the transformer state evaluation is carried out based on the Bi-LSTM network in different state evaluation intervals.The experiment proves the feasibility of analyzing the scientific nature of selecting the optimal time series and adding vibration index and winding hot spot temperature data to improve the accuracy of state evaluation.Based on the experimental results,the rationality and effectiveness of Bi-LSTM method for transformer state evaluation are proved.There are 48 figures,15 tables and 87 references in this thesis.